Envisioning Vocab

Vocab is a knowledge graph for all ML (Machine Learning) terms and concepts.

The best time to get into AI was 10 years ago. The second best time is now.

This will help you:

  • Follow the conversation and more easily understand recent developments in the field.
  • Advance your technological literacy and further your professional development.
  • Engage critically with the most important technology of our lifetimes.

How to use Envisioning AI:

  • Start with the basics and build up your conceptual knowledge.
  • Search for terms you are not familiar with.
  • Navigate the entire knowledge graph.

The entire index has been co-created with different AI tools in order to create a comprehensive overview of the field with accurate and useful definitions. If you spot a mistake or want to suggest missing concepts, email us on hello@envisioning.io.

Methodology

Definitions

  • Vocab articles are created by feeding each term into GPT-4o with instructions to create a comprehensive and accurate explanation of the term following a predefined style guide.

Generality

  • Each vocab term is assessed by GPT-4o in order to estimate its relative level of importance compared to the whole data set.
  • The language model evaluates each term several times, and an average score is presented here.

Year

  • We estimate the first citation or reference of each article with GPT-4o-mini and sometimes Claude 3.5.
  • These are manually compared against the time range indicated by article definitions.

Intersimilarity

  • The degree of similarity between each article pair is estimated in order to identify relationships between different concepts.
  • We use TF-IDF vectorization followed by cosine similarity.

Names

  • Selected with GPT-4o-mini.

Image

  • GPT4o: Receives title and summary,
  • Suggests image prompt for FLUX 1.1 Pro.

Fact Checking

  • LLMs have remarkable self-knowledge. This is likely because of the amount of research papers and articles used in the training data. This minimizes the chances of hallucinations or factual mistakes, which has been validated by running most entries through a fact-checking CustomGPT.

Visualizations

Graph

  • Generative knowledge graph of all articles. Experimental but fully functional
  • Connections are created based on node similarity, creating a semantic map of the corpus.

Grid

  • Experimentental quadrant view of metadata.

Learn more about ML

(cc) Envisioning